Judge Circuits Explain Format-Induced Inconsistency in LLM-as-a-Judge
LLM-as-a-judge has become the dominant paradigm for grading model outputs at scale, yet the same model assigns systematically different scores when its output format changes (e.g., a 1-5 rating vs. a True/False label). Existing diagnoses of these format-induced inconsistencies stop at the input-output level. Using Position-aware Edge Attribution Patching (PEAP), we causally investigate the internal mechanism in five open-weight instruction-tuned models (Gemma-3, Qwen2.5, Llama-3.1) across five judgment tasks. We find that judgments across structured understanding and open-ended preference tasks share a sparse Latent Evaluator sub-graph in the mid-to-late multi-layer perceptrons (MLPs); zero-ablating it collapses judgment while preserving performance on our knowledge probes in architecturally modular models. By structurally decoupling abstract judging from output formatting, we provide a mechanistic account of format-induced inconsistency on the open-weight models we study: a continuous judgment signal computed in the shared trunk is mapped through fragile, format-specific terminal branches. The judgment itself can therefore be read out independently of the requested output format. Our findings imply that benchmark comparisons of judge reliability across formats partly measure the fragile formatting stage, and can understate the quality of the underlying evaluation.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
Systematic Evaluation of LLM-as-a-Judge in LLM Alignment Tasks: Explainable Metrics and Diverse Prompt Templates
LLM-as-a-Judge has been widely applied to evaluate and compare different LLM alignmnet approaches (e.g., RLHF and DPO). However, concerns regarding its reliability have emerged, due to LLM judges' biases and inconsistent…
TrustJudge: Inconsistencies of LLM-as-a-Judge and How to Alleviate Them
The adoption of Large Language Models (LLMs) as automated evaluators (LLM-as-a-judge) has revealed critical inconsistencies in current evaluation frameworks. We identify two fundamental types of inconsistencies: (1) Scor…
Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models
We introduce methods for discovering and applying sparse feature circuits. These are causally implicated subnetworks of human-interpretable features for explaining language model behaviors. Circuits identified in prior w…
Language ModelingLanguage ModellingMCJudgeBench: A Benchmark for Constraint-Level Judge Evaluation in Multi-Constraint Instruction Following
Multi-constraint instruction following requires verifying whether a response satisfies multiple individual requirements, yet LLM judges are often assessed only through overall-response judgments. We introduce MCJudgeBenc…
Instruction FollowingEffects of cell cycle noise on excitable gene circuits
We assess the impact of cell cycle noise on gene circuit dynamics. For bistable genetic switches and excitable circuits, we find that transitions between metastable states most likely occur just after cell division and t…